Export Reviews, Discussions, Author Feedback and Meta-Reviews

This paper presents a recurrent convolutional neural network for semantic image segmentation to encode and take advantage of contextual relationships. The method is basically a combination of [13] where a very similar RCNN is used for object recognition, and [4] from which the multi-scale pipeline is inspired. Thus, technically, the paper is not very novel (Sec. However, the paper is well executed, very easy and clear to read, largely well written and provides a seemingly fair evaluation to state-of-the-art. Some recent works or interesting evaluations could be added, see below.